 |
Description  |
|
|
BACKGROUND OF THE INVENTION
This invention relates to methods and apparatuses for quickly measuring response and distortion characteristics of a signal transfer device, and in particular, for measuring the signal transfer characteristics of an audio signal transfer device,
such as an audio amplifier, a telephone channel, or the like.
In the sound reproduction, broadcast and telephone communications industries it is often necessary or desirable to determine the response and distortion characteristics of a signal transfer device in order to evaluate, repair or improve an audio
signal transfer path. The transfer device may be any of many commonly known devices, such as a stereo sound reproduction amplifier (sometimes referred to as a "hi-fi" or "stereo" amplifier), a telephone communications channel or other communications
link, a magnetic tape audio signal recorder, or audio signal broadcast equipment. What these devices have in common is that they accept as an input one or more audio frequency signals and reproduce them at their output, either immediately or delayed in
time.
The conventional way of measuring response and distortion characteristics of such devices ordinarily has been to apply sequentially one or more known test tones, i.e., audio frequency sinewave signals whose frequencies are known, to the input of
an audio signal transfer device and to measure their amplitude and relative phase, and the amplitude of their harmonics, at the output of the device. The tones are usually measured at the output using a bandpass filter whose center frequency is tuned to
one of the input frequencies or a harmonic thereof. For example, to determine the frequency response of such a device, a number of tones are applied sequentially to the input and their amplitudes are measured at the output to obtain data representative
of output amplitude as a function of frequency. Similarly, to measure harmonic distortion a known frequency is applied to the input of a device and the amplitudes of those harmonics of the input frequency which are present at the output are measured as
an indication of harmonic distortion. Conventional testing with manually-operated equipment can take as much as an hour for a thorough evaluation. This can be reduced with computer-controlled equipment, but still requires a significant amount of time
during which the device under test (hereinafter "DUT") cannot be used for normal activity. It is to be understood that the terms device or DUT used herein refer without limitation to one or more devices connected together.
A recognized international measurement standard has been adopted by the CCITT (International Consultative Committee on Telephone and Telegraph) and the EBU (the European Broadcasting Union) which employs two sequences of test tones, one for
monophonic devices and one for stereo devices. The monophonic sequence lasts 31 seconds and the stereophonic sequence lasts 33 seconds. The sequence begins with a preamble which indicates which sequence is being sent and the originator of the test.
The tones are one second in duration each, and an 8 second long pause is included for noise measurements. However, while this method has the advantage of being standardized, like other conventional test methods it is relatively time consuming; that is,
it requires the device under test to be shut down from normal activity for a significant amount of time, thereby disrupting normal operations.
There are also devices known as real time analyzers that apply a white noise signal to the input of a DUT and provide an indication of the output amplitude at a number of frequencies by the simultaneous use of multiple bandpass filters tuned to
those frequencies. The amplitudes of the outputs from those filters are displayed. However, those devices are limited in their usefulness in that they provide little information beyond frequency response, and they require time to average the noise
signal at the various frequencies measured.
Thence, it can be seen that there has been a need for a method and apparatus for making thorough signal response and distortion measurements of signal transfer devices quickly, so as to take them out of service for a minimal amount of time.
SUMMARY OF THE INVENTION
The limitations of conventional response and distortion measurement methods and apparatuses are overcome in the present invention by a form of digital parallel processing of the test signal input to a DUT and the resultant output. A plurality of
selected test tones are applied to the input of the DUT simultaneously. The output of the DUT is measured using a Fast Fourier Transform ("FFT") to convert the time domain output signal to the frequency domain, thereby permitting the amplitude, phase
and frequencies of the output signal components to be determined.
After the type of test to be performed and the test tones are selected, the input test signal is created in digital form by a computer processor and stored in a digital memory. When the test is to be run, the digitized test signal is read out of
memory and converted to analog form for input to the DUT, assuming that the DUT is not a digital audio signal processor. The output from the DUT is simultaneously acquired, converted to digital form (assuming that the device does not have a digital
output) and stored in the digital memory. In this way, the test signal application and acquisition time can be reduced to the period of just a few repetitions of the test signal stored in memory. If the DUT is a digital audio signal processor, no
digital to analog or analog to digital conversion is needed. A combination of analog and digital equipment may need only one digital to analog or analog to digital conversion.
Once the DUT output data is acquired, the DUT may be returned to normal operation, and computation of the FFT is performed to identify the amplitudes, phases and frequencies of the DUT output signal. Values for signal transfer characteristics
such as frequency response, harmonic distortion, intermodulation distortion, phase distortion, wow and flutter, and channel separation in a stereo or other multichannel system can be computed from test information displayed and produced in numerical form
or as a function of some parameter, such as frequency. In addition, the source of the test signal can be identified by the pattern, or signature, of the DUT output signal.
Accordingly, it is a principal objective of the present invention to provide a new and improved method and apparatus for measuring the signal transfer characteristics of a signal transfer device.
It is another objective of the present invention to provide a measurement method and apparatus that simultaneously applies a plurality of test tones to the device to be tested and transforms the output of the device to the frequency domain to
identify the amplitudes, phases and frequencies of the components that make up the output signal.
It is a further objective of the present invention to provide a measurement method and apparatus that digitally generates an input signal comprising a plurality of test tones simultaneously applied to a device to be tested and converts the output
of the device to digital form for analysis.
It is yet another objective of the present invention to provide a method and apparatus that permits the signal transfer characteristics of a signal transfer device to be determined in minimal time.
The foregoing and other objectives, features, and advantages of the invention will be more readily understood upon consideration of the following detailed description of the invention, taken in conjunction with the accompanying drawings.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 is a block diagram of an apparatus according to the present invention.
FIG. 2 is a schematic diagram of the basic principle of operation of the present invention.
FIG. 3A is a plot of a typical sinusoidal waveform used in conventional response and distortion measurements.
FIG. 3B is a plot of a typical DUT input waveform used in the present invention on the same relative amplitude scale as FIG. 3A.
FIG. 4 is an illustration of three tone signal waveforms relative to a signal generation time block during which an output signal is acquired, a first one having a period equal to that time block and two others having periods that are integer
multiples of the frequency of the first one.
FIG. 5 is an illustration of a plurality of frequency subdivisions, or "bins," corresponding to the allowable frequencies that can be measured at the output of a DUT when using a Fast Fourier Transform according to the present invention.
FIG. 6A is a plot of input test tone amplitudes as a function of frequency for a sixty tone test signal.
FIG. 6B is a plot of illustrative measured DUT output amplitudes as a function of frequency for a sixty tone test signal.
FIG. 6C is a plot of input test tone amplitudes as a function of frequency for a sixty tone test signal, weighted to simulate program material.
FIG. 7 is a plot of illustrative DUT output amplitudes as a function of frequency when excited with a single sinewave, showing harmonic distortion components to be measured in accordance with the present invention.
FIG. 8 is a plot of illustrative measured DUT output amplitudes, at harmonic frequencies only, as a function of frequency for a five tone test signal.
FIG. 9 is a plot of input test tone amplitudes as a function of frequency for a five tone test signal.
FIG. 10 is a plot of illustrative DUT output amplitudes as a function of frequency, simplified to show only two test tone intermodulation and wow and flutter components to be measured in accordance with the present invention.
FIG. 11 is a plot of illustrative measured DUT output amplitudes, at intermodulation frequencies only, as a function of frequency for a five tone test signal.
FIG. 12 is a plot of an illustrative DUT input signal amplitude as a function of time, for a sixty tone test signal with randomly selected phase.
FIG. 13 is a plot of illustrative measured DUT output amplitudes, at selected unused frequencies, as a function of frequency for a five tone test signal.
FIG. 14A is an illustration of a DUT output waveform whose period has been lengthened by the DUT.
FIG. 14B is an illustration of window waveform.
FIG. 14C is an illustration of the product of the window waveform shown in FIG. 14B with the waveform shown in FIG. 14A, as found in accordance with the present invention.
FIG. 15 is a plot of an illustrative DUT output amplitude as a function of time for a multi-tone test signal, where one test tone has an amplitude about 12 dB higher than the amplitudes of the other test tones.
FIG. 16A is a plot of illustrative measured DUT output amplitude as a function of frequency for a five tone test signal input, showing frequency bands, not including the test tones, that are to be root square summed.
FIG. 16B is a plot of the root square sum values computed for the bands (I, IiI, III, IV, V and VI) iranges illustrated in FIG. 16A as a function of the center frequency of the band.
FIG. 17A illustrates the identification of the presence of a pattern of frequency components in the output of a DUT having amplitudes exceeding predetermined required component minimum amplitude thresholds.
FIG. 17B illustrates the identification of extra frequency components in the output of a DUT having amplitudes exceeding a predetermined maximum threshold.
DETAILED DESCRIPTION OF THE INVENTION
A preferred embodiment of the present invention and its relationship with a device under test ("DUT"), is shown in FIG. 1. An apparatus according to the present invention preferably comprises an input memory 10 and an output memory 12, both of
which preferably are random access memories which are part of a computer processor 14, and may be subdivisions of the same memory. The computer processor includes a central processing unit 16 (hereinafter "CPU") which is interconnected with memories 10
and 12 and programmed to carry out the functions of the present invention.
Under control of the CPU 16, the input memory 10 supplies to a digital analog converter 18 stored data representing a digitized waveform to be provided to the input of the DUT 20. The waveform is predetermined by the CPU in accordance with user
specifications of amplitude, phase and frequency of the test tones, and stored in the input memory 10 for quick access during testing of the DUT. A digital to analog converter 18 converts the digitized waveform supplied by the input memory 10 to an
analog signal for application to the input 22 of the DUT. The output 24 of the DUT is supplied to analog to digital converter 26 which digitizes that analog signal and stores it in output memory 12. The results of test computations made on that stored,
digitized output data are displayed on a video monitor 28 or other suitable alphanumeric or graphical display device. This enables the test data to be acquired in minimal time so that the DUT can be returned to service in a negligible period of time,
e.g., one second, after which the processing explained hereafter can take place. At the same time, it is to be recognized that the method and apparatus of the present invention as described herein could be implemented with other hardware and that the
device under test could be a digital audio signal processor which would not require the digital to analog converter 18 or the analog to digital converter 26, or both, and that it could have a digital input with an analog output or an analog input with a
digital output, without departing from the principles of this invention.
The present invention employs a type of parallel processing by applying to the DUT 20 a plurality of sinusoidal test tones at once, as represented by f.sub.1, f.sub.2, f.sub.3, and so on, up to f.sub.i, in FIG. 2, where i is the number of
distinct test tones applied to the device under test at once. These tones are added, as represented by symbol 30, and applied to the input 22 of the DUT 20. The output 24 of the DUT is measured and analyzed to determine the spectral characteristics,
that is, the frequencies, phases and amplitudes of those frequencies which are components of the output waveform of the DUT. Those frequencies, which may be fewer, but generall is greater than the number of input frequencies, are represented by
w.sub.1,w.sub.2, w.sub.3 and w.sub.o in FIG. 2, where o represents the number of distinct frequencies found in the output waveform. By analyzing the frequencies, phases and amplitudes of the components that make up the output signal it can be determined
whether there is harmonic distortion, intermodulation distortion, amplitude distortion, or phase distortion. The amount of noise present, the existence of wow and flutter, a change in speed of the signal as, for example, where a tape recorder is running
too slow or too fast, and in the case of a multichannel system, the degree of separation of channels, can also be determined. The computer processor 14 may be programmed in any convenient way to perform the measurements set forth herein.
While the type of signal typically applied to the input of a DUT in the prior art ordinarily comprises one, or perhaps two, sinusoidal signals as shown in FIG. 3A, the signal applied to the device under test in the present invention would
ordinarily appear more or less like that shown in FIG. 3B, depending upon the specific frequencies, phases and amplitudes of the test signal frequencies f.sub.1 through f.sub.i.
Turning now to FIG. 4, the period T represents the time during which the input signal 22 to the DUT 20 completes one waveform repetition, and an integer sub-multiple of the time during which the output signal 24 of the DUT 20 is acquired in the
method and apparatus of the present invention. That is referred to hereinafter as a signal generation time block or "generation block." Waveform 32 represents a generation block repetition frequency having the period T. Waveforms 34 and 36 represent two
test tones, though there may be many more. The generation block repetition frequency is both the lowest frequency which can be generated and the closest possible spacing of the test tones. Each of the test tones is an integral multiple of the
generation block repetition frequency; for example, test tone 34 is two times the generation block repetition frequency 32, and test tone 36 is five times the block repetition frequency 32. However, it is desirable that the test tones, and their
harmonics, not be integral multiples of one another, because that would make identification of harmonic distortion at the output of the DUT more difficult.
The output measurement and analysis is accomplished preferably by sampling the DUT output in the time domain at a frequency at least twice as high as the highest frequency component to be measured at the output of the device under test, as
dictated by the sampling theorem. As indicated above, the sample data is digitized, stored and operated on by the CPU 16. The preferred technique to be used is a Fast Fourier Transform ("FFT") which converts the time domain data to the frequency
domain. However, other time domain to frequency domain transforms might also be used without departing from the principles of the invention. In the preferred embodiment of the invention, the DUT output signal is sampled at the same rate a the test
signal samples applied to the output. This causes all frequency components in teh signals to fall in the center of measurement FFT frequency bins. This technique is well known in the electrical engineering art, and many useful computer algorithms can
be found to implement the technique.
The output is measured over a period referred to hereinafter as a measurement time block or "measurement block." The measurement block is an integer multiple of the generation block. The generation block period T is typically chosen so that the
measurement block has a number of data samples equal to a power of 2, though this is not required for some FFT algorithms. The FFT yields one distinct measurable frequency for each pair of input samples. The minimum frequency resolution is equal to the
sample rate divided by the number of distinct measurable frequencies. For example, for 16,384 input samples at a sample rate of 48,000 Hz, there will be 8,192 distinct measurable frequencies with a spacing of 2.93 Hz. Preferably the lowest test tone
for most audio systems would be 20 Hz, or in the case of the aforementioned example, 17.58 Hz.
Since a fixed number of points i in the time domain are acquired for analysis, the test tones may be generated digitally to exactly match the measurement block. Also, the frequency spectrum available to be measured at the output of the device
under test may be conceptually divided into a finite number of frequencies that can be separately identified, referred to hereinafter as "bins" as shown at 38 in FIG. 5. Each bin is separated by the frequency equal to the highest resolution of the
system. For example, if the block repetition frequency is 2.93 Hz, the center of each bin will be separated by that amount.
A commonly known characteristic of an audio transfer device is its frequency response. This is the variation of device gain with frequency. In the present invention, rather than applying a single frequency to the input of the DUT and measuring
its amplitude at the output and repeating the process for many different input frequencies, a plurality of frequencies are simultaneously applied to the input, and their corresponding amplitudes at the output are identified by the FFT. In the frequency
domain the input would be as shown in FIG. 6A, where the amplitude is plotted as a function of frequency, on a log scale. In this exemplary case, a 60 tone input is employed, though fewer or greater numbers of input tones, or frequencies, may be
employed. The amplitude of the output is measured only at the input frequencies, and plotted as a function of frequency, as shown in FIG. 6B. In FIG. 6B the amplitudes of frequencies between those corresponding to the test tones have been interpolated
in order to obtain a continuous curve. While interpolation is not necessary, it is desirable because it makes the output easier to read and estimation of amplitudes between the distinct test tones easier.
In order to measure the frequency response of the DUT under nearly actual operating conditions, the input amplitudes may be weighted as a function of frequency to simulate frequency amplitudes in program material, as shown in FIG. 6C.
Wow and flutter, a type of frequency modulation distortion, is often produced by audio tape players. This has the effect of shifting energy from test tones into side bands of the test tones. The amplitude measurement of the acquired test tone
frequency will produce an incorrect value; that is, it will be too low. This problem may be solved by root sum square addition of the tone frequency and its neighboring sidebands.
Another characteristic of audio signal transfer devices is the extent of harmonic distortion that they produce. In FIG. 7 amplitude is plotted as a function of frequency. An input signal having frequency f is shown, along with its second,
third, fourth and highest identified harmonics, w.sub.1,w.sub.2, w.sub.3 and w.sub.n. While this represents the output that would be obtained by a single test signal where the device under test produces substantial harmonic distortion, a multitone test
signal may be employed with the present invention to obtain a comprehensive harmonic distortion measurement in a short period of time. For example, FIG. 8 shows a plot of output harmonic amplitude as a function of frequency where a five tone test
signal, as shown in FIG. 9, has been applied to the input of the DUT. The five test tones are selected to not be integer multiples of each other so that the harmonics of those tones will occur at unique places in the spectrum. A total harmonic
distortion (hereinafter "THD") value can be obtained by taking the root sum square of the measured harmonic distortion component output amplitudes; that is: ##EQU1## where H=the number of harmonic components, and A.sub.H =the amplitude of a given
harmonic component.
If test tones were chosen so that the harmonics were allowed to overlap, it would be possible for harmonics with opposing phases to cancel, thereby producing lowered, incorrect distortion readings. If the DUT has a nonlinearity with a sharp
discontinuity, it will generate high order harmonics. Low frequency test signals will then have harmonics which extend all across the audio band. This makes selection of test tones difficult if harmonic overlap is to be avoided. However, most
nonlinearities produce a distortion spectrum which falls off with increasing harmonic order. This allows an upper limit on harmonic order to be used to calculate which test tones will cause overlap. Above that limit the amplitude of harmonics are
likely to be negligible when compared with the components already included in the output. In FIG. 8 the amplitudes between the various harmonic frequencies actually measured are interpolated to provide a continuous curve.
Sometimes it is desirable to make a measurement of the harmonic distortion performance of a device which can be compared to conventional harmonic distortion measurements, yet the fast frequency response and separation measurement capabilities of
the present invention are still needed. Raising the amplitude of one test signal frequency component significantly above the others will cause the waveshape and crest factor to be dominated by the larger tone. The harmonics created by the DUT when
excited by this signal will be dominated by the harmonics of the larger tone. In practice, an amplitude difference of about 12 dB is sufficient to provide reasonable correlation with single sinewave based measurements in most cases, although some have
required a level ratio as high as 30 dB. Such a waveform is illustrated in FIG. 15. When this technique is used, the other test tones may still be used to obtain all other measurement capabilities of the present invention, such as frequency response
and channel separation measurements.
Measurement of intermodulation distortion (hereinafter "IMD"), and wow and flutter in the present invention is explained with respect to FIG. 10. In FIG. 10 one frequency of a multitone input signal is shown by f.sub.b and another, higher
frequency test tone is represented by f.sub.a. Ordinarily there would be more than just those two test tones, but FIG. 10 has been limited to those two tones for purpose of clarity. The higher IMD components are represented by (f.sub.a +f.sub.b), or
"f.sub.a+b ", and (f.sub.a +2f.sub.b), or "f.sub.a+2b ". Similarly, the lower intermodulation distortion components are represented by (f.sub.a-b), or "f.sub.a-b ", and (f.sub.a-2fb), or "f.sub.a-2b ". That is, intermodulation between f.sub.b and
f.sub.a produces sum and difference frequencies in the output. It is recognized that more distortion products between those two tones may be generated, but only these are shown for the purpose of explanation. Wow and flutter, a type of frequency
modulation distortion, is often produced by audio tape players. That also produces distortion components, shown in FIG. 10 as (f.sub.a +f.sub.c), or "f.sub.a+c "; (f.sub.a -f.sub.c), or "f.sub.a-c "; (f.sub.a +f.sub.d), or "f.sub.a-d," where f.sub.c and
f.sub.d represent wow and flutter frequency components.
If the DUT has nonlinearities, there will be IMD components between all combinations of tone frequencies. By modeling the nonlinearity as a power series, the IMD frequencies can be predicted. These IMD components will appear at frequencies
above, below and between the test tones. The calculation of these IMD frequencies becomes very complex when many test tones are included and when the nonlinearity is of a high order. Thence, selection of the test tones must be done carefully.
In applications involving large numbers of test tones, i.e., about 6 or more, the distortion spectrum will become complex enough that although some components may cancel there will be a sufficient number of distortion components remaining that an
accurate assessment of nonlinearity may still be made. However, the sheer number of distortion components involved may make the graphing time excessive or the interpretation difficult. By adding all components except the original test tones themselves,
a measurement is made which does not suffer from these limitations.
To allow correct summation, a root-sum-square (often called root square summation) technique is used to produce an indication of the effective level in the distortion components. If the measurements are performed using the results of an FFT this
merely involves root square summation of the bin levels for all bins excluding those of the original tones. The fact that frequencies are included which may not have distortion components will cause the effects of noise to be included in the
measurement. This makes the measurement analogous to the total harmonic distortion plus noise (THD+N) measurement which is commonly performed when using a single sinewave stimulus. Since a multiple sinewave stimulus is used, the measurement will also
include intermodulation products and is most accurately called total distortion and noise. If some indication of the distortion variation with frequency is desired, the summation may be done in parts, or frequency bands, as illustrated in FIG. 16A, each
band covering a portion of the frequency range, and the sums graphed as a function of the center frequency of that band, as shown in FIG. 16B.
In cases where wow and flutter is not a significant error source in the measurement, a numerical representation of IMD may be obtained by root square sum addition of the components at the intermodulation frequencies. A visual display of IMD is
obtained by plotting the amplitudes of the output at IMD frequencies only, as a function of frequency, as shown in FIG. 11. The amplitudes of frequencies between the IMD components are interpolated to provide a continuous curve.
In order to obtain a numerical value representative of IMD when wow and flutter is present, bins located symmetrically on either side of a test tone are summed linearly in pairs, but with a phase inversion of one of the frequency components in
each pair. Then the pairs are root sum squared; that is: ##EQU2## where N=the number of sideband pairs, and A.sub.UN =the upper sideband, and
A.sub.LN =the lower sideband.
The inner subtractionis of the vector amplitudes which include phase information. This produces a value for IMD sidebands, but rejects wow and flutter sidebands in the spectrum measured. This is because the upper and lower wow and flutter
sidebands are in phase (so they cancel out when one is inverted), while the upper and lower IMD sidebands are 180.degree. apart in phase (so they add when one is inverted).
To obtain a wow and flutter distortion numerical value, the analysis bins located symmetrically about a test tone are summed linearly in pairs (without inversion of one). Then the pairs are root square summed, thereby canceling out the IMD
sidebands in the spectrum measured; that is: ##EQU3## where N=the number of sideband pairs, and A.sub.UN =the upper sideband, and
A.sub.LN =the lower sideband.
The inner addition is of the vector amplitudes which include pahse information.
The distortion measurements obtained by this technique are not directly comparable to those obtained by single sinewave total harmonic distortion plus noise ("THD+N") testing, or by conventional IMD testing. This is because the crest factor,
i.e., the ratio of peak to RMS amplitude as illustrated in FIGS. 3A and 3B, of a multitone input signal, will always be higher than that of a single sinewave; that is, for the same peak signal amplitude, the amplitude of each individual tone will be
lower than a single sinewave at that tone. However, an accurate and reliable result may be obtained by this technique to provide good comparison of DUT characteristics over time, or to other devices tested by the same technique. It may be desirable to
scale distortion measurements by a correction factor to yield readings which are comparable to that obtained with conventional testing for typical DUTs. This can be done by measuring similar DUTs in the conventional way and deriving correction factors
by comparing the two sets of measurements.
In order to minimize the crest factor, and thereby reduce the likelihood of clipping distortion, the test tone signals may be provided with random phases, as shown in FIG. 12 for a sixty tone test signal. Alternatively, the crest factor can be
made similar to that of program material by fixing the phase of each of the sinewave components of the input signal to create a test signal with the desired crest factor. This is because, since all the test tones are a multiple of the generation block
repetition frequency, their phases are locked together.
Noise produced by the DUT may be measured by selecting test tones which leave gaps in the harmonic and intermodulation distortion component spectrum, and measuring the amplitude of the output signal at frequency bins between the test tones and
their harmonic and IMD frequencies. A plot of the amplitude values obtained by this technique, as a function of frequency, with interpolation between distinct frequencies, is shown in FIG. 13. A numerical noise value measurement can be obtained by root
square summing the amplitudes obtained this way. However, the squared and summed value must be multiplied by a constant representing the number of bins used in the computation, the bandwidth of the bins, and the bandwidth of the measurement to yield an
accurate wide band noise figure. The computation is made as follows: ##EQU4## where N=number of bins used in noise calculation, A.sub.I =amplitude in bin number N,
N.sub.B =number of bins across the entire output measurement spectrum; and
C.sub.WG =window gain correction constant, as commonly known in the use of FFTs.
If enough frequency points are used it is possible to factor in weighting filter gain as a function of frequency when computing noise, thereby yielding a weighted noise measurement. This is done by multiplying each A.sub.I by the weighting
filter gain at each bin frequency before the squaring and summation in the above equation.
When using signals with a large number of test frequencies it may be difficult to find empty bin frequencies. By making the measurement block length an integer multiple larger than one of the generation block length, there will be additional
resolution in the measurement spectrum. This will ensure empty bin frequencies since the generated frequencies, and the resulting harmonic and intermodulation frequencies, must always be an integer multiple of the generation block repetition rate. By
making the measurement block length twice that of the generation block length, every alternate frequency bin will be guaranteed to be free of distortion products. Consequently, summing only the alternate bins will result in a noise measurement
uncorrupted by distortion.
Applying different test signals to two channels of a stereo DUT allows measurement of | | |
